Trimmed mean trade data cuts US forecast errors up to 15 percent
New trimmed import and export growth rates for the United States reduce forecast errors by up to 15 percent over four to 12 months. Federal Reserve Bank of Kansas City researchers Thomas Cook, Mariia Dzholos, and Johannes Matschke introduced the publicly available methodology in August 2026.
Filtering the noise from volatile trade data
Authors Thomas R. Cook, Mariia Dzholos, and Johannes Matschke introduce trimmed import and export growth rates to address the notorious volatility of US trade flows.
The framework calculates growth rates as a trimmed mean, systematically stripping out volatile subcomponents with weak predictive value to isolate underlying trends.
Out-of-sample tests show the trimmed indicator cuts aggregate trade forecast errors by up to 15 percent, delivering the largest gains at horizons between four and 12 months ahead.
The authors note that the simple metric reliably outperforms naive projections as well as complex models using broader datasets, particularly during periods of elevated trade volatility.
Adapting core inflation tools to trade
The methodology adapts the trimmed mean approach previously applied to consumer price index and personal consumption expenditures inflation to international trade flows.
Published as Kansas City Fed Research Working Paper 26-08, the series builds on prior trade forecasting research by Cook, Lusompa, and Matschke.
The researchers have made the monthly dataset publicly accessible to provide policymakers and forecasters with cleaner real-time indicators.
Useful filter, incomplete picture
The paper delivers a practical forecasting upgrade by borrowing a proven inflation measurement concept.
However, discarding volatile components risks overlooking structural realignments and sharp tariff adjustments during geopolitical shifts.
For trade analysts, it remains a useful benchmark rather than a standalone forecasting solution.
Source: Introducing Trimmed Imports and Exports
IN: